How to assess the bandwith required on the proxy for Discovery traffic between MID server and instance?

[Marek Meres]

Hello Community,

I would like to assess the bandwidth on the proxy that will be consumed by my planned MID servers (we are at the design phase) to contact ServiceNow instance when running Discovery.

We will have 5 MID servers scanning about 600 Windows/Linux servers mainly, plus a couple of network switches and routers, Discovery on default setting (collecting all possible information).

On the https://docs.servicenow.com/bundle/rome-it-operations-management/page/product/discovery/reference/discovery-resource-utilization.html I can see discovering one server with patterns causes 0.3-0.4 MB of data to be send to the instance and 0.4-0.5 from the instance (it seems the patterns are bigger than the results then). I guess I need to add to this around 0.6-0.7 per "3 tier application" on the server, let's assume around 1 MB then for complete server with app on it. But this is total amount of bytes transferred over the whole discovery time which my be several minutes as per my experience... And with multiple servers being scanned on various stages in the same moment in time by different MID servers all using the same proxy it becomes problematic (?) to assess how much bandwidth the traffic will consume on the proxy (which is my goal here).

Do you have any experience on that? I am of course not asking or accurate numbers, just a rough assessment - would it be 10 Mbps, 100 Mbps etc? In my situation these are 5 MID servers scanning roughly 50-200 servers each but any numbers with relation to the different configuration would be much appreciated!

Thank you in advance!

Marek

Many thanks in advance

[Fabian Kunzke]

Hello,

We have run tests like these with multiple customers and never found an issue. What we did was fearly simple:

On every mid-server we ran a wireshark (or equivalent) network listener on the most commonly used discovery ports (22 for ssh, 135 for wmi, 161 for SNMP). We then started a discovery.

We repeated that procedure for some of the servers we discovered. We setup a wireshark listener there as well.

This gave us 2 results: 1 for the Mid Server and 1 for the discovered hosts. This allows us to very accurately determine what kind of load is to be expected.

Note: Make sure to include different sized servers for these tests. Also, some firewalls allow you to track this traffic load as well, making the need of wireshark irrelevant.

Our results were mostly underwhelming. I would highly doubt that you manage to load your network with 100Mbps when discovering. Remember: A mid server can not discover many devices at once (i couldn't find the exact number). This is also why one of the simplest ways to cut down on discovery time is by deploying another mid server.

Feel free to let me know if you need any more details.

Regards

Fabian

[Marek Meres]

Hello Fabian,

Thank you! I am concerned about the proxy that will forward the traffic from all 5 MID servers to/from the SN instance. Do you think the 100 Mbps boundary is enough for that traffic as well? My problem is that I need to assess this now, even before the network is built (this is a new environment)...

Best regards

Marek

[Fabian Kunzke]

Hey Marek,

I cannot vouch for it being enough, but i would highly doubt it. If you have the chance to at least test it with one MidServer, please do so. This will give you a clear number.

In short: I personally never witnessed a mid server exceeding that kind of data limit. However, i can not ensure you. If you need to be sure, at least try to simulate the discovery with one mid server and a sandbox enviroment (a standalone ServiceNow instance).

Regards

Fabian